Frailty markers comprise blood metabolites involved in antioxidation, cognition, and mobility

Masahiro Kamedaa, Takayuki Teruyab, Mitsuhiro Yanagidab,1, and Hiroshi Kondoha,1

aGeriatric Unit, Graduate School of Medicine, Kyoto University, Sakyo-ku, 606-8507 Kyoto, Japan; and bG0 Cell Unit, Okinawa Institute of Science and Technology Graduate University, Onna-son, 904-0495 Okinawa, Japan

Contributed by Mitsuhiro Yanagida, March 3, 2020 (sent for review December 2, 2019; reviewed by Hidenori Arai and Elizabeth H. Blackburn) As human society ages globally, age-related disorders are becom- aging) is affected not only by age, but also by disease, psycho- ing increasingly common. Due to decreasing physiological reserves physiological condition, and lifestyle (7, 8). and increasing organ system dysfunction associated with age, Metabolomics, a tool for evaluating metabolite profiles, em- frailty affects many elderly people, compromising their ability to ploys liquid chromatography–mass spectrometry (LC-MS) to cope with acute stressors. Frail elderly people commonly manifest reveal complex but highly integrated biological processes (5). complex clinical symptoms, including cognitive dysfunction, hypo- Although noncellular components (serum or plasma) of blood mobility, and impaired daily activity, the metabolic basis of which have most often been used for metabolomic assays (5), we de- remains poorly understood. We applied untargeted, comprehen- veloped whole blood and metabolomics (9) to sive LC-MS metabolomic analysis to human blood from 19 frail and comprehensively investigate metabolic foundations of human nonfrail elderly patients who were clinically evaluated using the aging. Metabolomic analyses enable us to detect metabolites Edmonton Frail Scale, the MoCA-J for cognition, and the TUG for related to metabolism, the tricarboxylic acid (TCA) mobility. Among 131 metabolites assayed, we identified 22 cycle, nitrogen, sugar, purine/pyrimidine, lipid metabolism, markers for frailty, cognition, and hypomobility, most of which antioxidation, energy supply, and diet. were abundant in blood. Frailty markers included 5 of 6 markers Based on these quantitative, reproducible analytical methods, specifically related to cognition and 6 of 12 markers associated we recently reported 14 age-related metabolites relevant to with hypomobility. These overlapping sets of markers included antioxidative defense and nitrogen metabolism (10). Four recent metabolites related to antioxidation, muscle or nitrogen metabo- reports drew divergent, nonoverlapping conclusions (11–14), lism, and amino acids, most of which are decreased in frail elderly stemming from different experimental designs. For example, the MEDICAL SCIENCES people. Five frailty-related metabolites that decreased—1,5-anhy- former two reports applied the Fried CHS index as a diagnostic droglucitol, acetyl-carnosine, ophthalmic acid, leucine, and isoleu- tool, efficiently detecting hypomobility but offering no cognitive cine—have been previously reported as markers of aging, assessment (2), while the latter two used a 70-item clinical Frailty providing a metabolic link between human aging and frailty. Index focusing mainly on activities of daily living (ADL) as- Our findings clearly indicate that metabolite profiles efficiently sessment (15). Here we report the untargeted metabolomic distinguish frailty from nonfrailty. Importantly, the ergothioneine, which decreases in frailty, is neuroprotective. Oxi- Significance dative stress resulting from diminished antioxidant levels could be a key vulnerability for the pathogenesis of frailty, exacerbating illnesses related to human aging. Frailty resulting from age-related deterioration of multiple or- gan systems displays complex features, including cognitive dysfunction, hypomobility, and impaired daily activity. How- frailty | | cognitive impairment | metabolomics | aging marker ever, metabolic aspects of frailty remain unclear. We per- formed untargeted, comprehensive metabolomics of whole uman society is aging globally, in developed as well as in blood from 19 frail and nonfrail elderly patients. We identified Hdeveloping countries, and people over age 85 now constitute 22 markers, including 15 for frailty, 6 for cognition, and 12 for 1.6% of the world population. While life expectancy is in- hypomobility, most of which are abundant in blood. Frailty creasing, there is also an alarming rise in the number of frail markers include 5 of 6 for cognition and 6 of 12 for hypo- people who are predisposed to be bedridden and to require mobility. These overlapping markers include decreased levels nursing care. The prevalence of frailty among those aged 65 and of metabolites related to antioxidation, nitrogen, and amino over is estimated at 17%, or approximately 120 million individ- acid metabolism. Ergothioneine, an antioxidant involved in uals worldwide (1). Frail people suffer not only from physical neuronal diseases, declines in frailty. Thus, we reveal essential disabilities, but also from psychophysiological and social prob- metabolites linked to the pathogenesis of frailty, including lems (2), and thus require more social resources than healthy vulnerability to . peers. Frailty compromises their ability to cope with acute Author contributions: M.Y. and H.K. designed research; M.K., T.T., M.Y., and H.K. per- stressors due to declining physiological reserves and organ sys- formed research; M.K. and T.T. analyzed data; and M.K., T.T., M.Y., and H.K. wrote tem function (2, 3), although it has been suggested that frailty the paper. may be reversible (4). Moreover, human aging is a highly com- Reviewers: H.A., National Center for Geriatrics and Gerontology; and E.H.B., University of plex biological process exhibiting great individual variation, and California San Francisco Medical Center. until now, its metabolic basis has been little understood. The authors declare no competing interest. Because all tissues and organs are supplied by the circulatory This open access article is distributed under Creative Commons Attribution-NonCommercial- system, blood should reflect environmental conditions, genetic NoDerivatives License 4.0 (CC BY-NC-ND). and epigenetic factors, nutritional status, exposure to exogenous Data deposition: Raw LC-MS data in mzML format are available from the MetaboLights repository, http://www.ebi.ac.uk/metabolights/. substances, and lifestyle factors (5, 6). Therefore, human blood 1To whom correspondence may be addressed. Email: [email protected] or hkondoh@ samples are expected to document not only individual genetic kuhp.kyoto-u.ac.jp. variability, but also differences in physiological responses and This article contains supporting information online at https://www.pnas.org/lookup/suppl/ homeostatic mechanisms. For example, recent studies suggest doi:10.1073/pnas.1920795117/-/DCSupplemental. that in circulating leukocytes, telomere length (a biomarker of First published April 15, 2020.

www.pnas.org/cgi/doi/10.1073/pnas.1920795117 PNAS | April 28, 2020 | vol. 117 | no. 17 | 9483–9489 Downloaded by guest on September 24, 2021 analysis of blood from frail and nonfrail elderly people. For mild cognitive impairment (10). Clinical attributes of the study frailty diagnosis, we applied the Edmonton Frail Scale (EFS) and participants are summarized in SI Appendix, Table S1. the Japanese version of the Montreal Cognitive Assessment First, we clinically evaluated whether the 19 participants were (MoCA-J) to evaluate cognitive aspects of frailty (16, 17). We frail, cognitively impaired, or hypomobile, according to EFS, show that antioxidants, amino acids, and metabolites related to MoCA-J, and TUG scores. Nine individuals (average age, 88.2 ± muscle or nitrogen metabolism link frailty to cognitive impairment 6.8 y) were diagnosed as frail (average EFS, 9.0 ± 1.2), while 10 and hypomobility. (average age, 80.5 ± 4.7 y) were not (average EFS, 4.7 ± 1.1) (SI Appendix, Table S1). According to the MoCA-J assessment, 15 Results and Discussion individuals displayed impaired cognition (average score; 19.3 ± Nineteen elderly participants, including 7 males and 12 females 3.8), while 4 were normal (average 27.0 ± 0.8). Regarding mo- with a mean age of 84.2 ± 6.9 y, were examined using the EFS, bility, 12 participants exhibited a prolonged TUG test (>10 s), the MoCA-J for cognitive function (17, 18), and the Timed Up & while 7 were normal (SI Appendix, Table S1). Both the MoCA-J Go (TUG) test for motor ability (19) (Fig. 1A). The EFS is an and TUG results were significantly diminished in frailty (Fig. 1B efficient diagnostic tool, comprising 10 questions to assess cog- and SI Appendix, Table S1). Significant correlations of EFS with nitive ability (clock-drawing test), mobility (TUG test), and MoCA-J, TUG, and functional independence test results fundamental daily activity, in which a score ≥7 indicates frailty (Fig. 1C and SI Appendix, Fig. S1) confirm that frailty involves (on a scale of 0 to 17) (16). It also covers domains related to simultaneously deteriorating physiological functions and social health status, functional independence, social support, medica- activities. tions, nutrition, mood, continence, and illness burden. The In this context, we performed untargeted analysis of 131 MoCA-J evaluates short-term memory, visuospatial ability, var- compounds in whole blood (Dataset S1). Our comprehensive ious executive functions, attention, concentration, working comparison of these metabolites between frail and nonfrail el- memory, language, and temporal and spatial orientation. An derly identified 15 compounds as frailty markers (Fig. 1D and SI MoCA-J score below a threshold of 25 to 26 (out of 30) indicates Appendix, Table S2). In addition, we found that 6 metabolites

Fig. 1. Metabolomic study of frailty. (A) Diagram of the study protocol. All participants were clinically examined, and their blood was analyzed using untargeted comprehensive metabolomics. (B) Comparison of MoCA-J and TUG test results between frail and nonfrail subjects. **P < 0.01. Error bars represent mean ± SD. (C) Pearson’s correlation of the linear model between EFS and MoCA-J (Left) or TUG (Right). (D) Overview of identified metabolites related to EFS, MoCA-J, and TUG.

9484 | www.pnas.org/cgi/doi/10.1073/pnas.1920795117 Kameda et al. Downloaded by guest on September 24, 2021 involved in cognitive impairment and 12 metabolites related to Pearson’s correlation analysis revealed a close relationship of TUG low mobility were also significantly changed (Fig. 1D). results with three metabolites: isovaleryl-carnitine, adenine, and Among the 15 frailty markers, 13 compounds—acetyl-carnosine, UDP-glucuronate) (SI Appendix,TableS4). N3-methyl- is ergothioneine (ET), S-methyl-ergothioneine (S-methyl-ET), trimethyl- an indicator of muscle deterioration (28), while isovaleryl-carnitine, histidine (hercynine), ophthalmic acid (OA), 2-ketobutyrate, enriched in muscle, supplies acetyl-CoA to mitochondria. Hippu- urate, 1,5-anhydroglucitol (1,5-AG), proline, isoleucine, leucine, rate, synthesized in mitochondria from metabolized poly- tryptophan, and methionine—decreased in frailty, while two metab- phenol, is involved in nitrogen metabolism (29). Arginine is also olites enriched in red blood cells—creatine and UDP-glucuronate— utilized in the urea cycle. These four hypomobility metabolites are increased. Ten of 15 frailty markers showed correlations with EFS involved in muscle or nitrogen metabolism, in addition to the in- scores (SI Appendix,TableS2). Based on the MoCA-J, a comparison volvement of three muscle-related amino acids—tryptophan, iso- of cognitively impaired subjects and controls detected significant leucine, and leucine—in frailty (Figs. 2B and 3C). changes in six metabolites—acetyl-carnosine, ET, tryptophan, Women are at an intrinsic increased risk of frailty by virtue of creatine, UDP-glucuronate, and UDP-glucose—all of which except lower lean mass and strength compared with age-matched men UDP-glucose were also frailty markers (Fig. 1D and SI Appendix, (2). There were significant differences in the skeletal muscle Table S3). Among these six metabolites, three compounds—tryptophan, mass index between males and females in our cohort (average, creatine, and UDP-glucuronate—displayed correlations with the 7.49 vs 5.31; P = 0.00003), while metabolites related to muscle MoCA-J results (SI Appendix, Table S3). Thus, Pearson’s cor- mass include compounds involved in nitrogen metabolism (30). relation of these marker metabolites also disclosed their con- Such compounds could be affected in sarcopenic frailty, with an gruence with relevant clinical attributes (SI Appendix, Fig. S2). inherent risk in women. We previously reported 14 aging-related markers (10). In the In addition to increased N-acetyl-aspartate in hypomobility, present study, we noticed that five frailty-related metabolites that three up-regulated metabolites were identified: creatine and were decreased—acetyl-carnosine, OA, isoleucine, leucine, and UDP-glucuronate for frailty and UDP-glucose for cognition 1,5-AG—were also among these aging markers (Fig. 2 and SI (Fig. 3D). Creatine, which is increased in cognitive impairment Appendix, Fig. S3A), insinuating a metabolic connection between and frailty, serves as major energy storage in brain and muscle frailty and human aging. (31). Creatine supplementation is effective for strengthening Strikingly, among 15 frailty markers, 7 compounds that were muscle in athletes (32) and for treating some types of genetic decreased are relevant to antioxidative defense: acetyl-carnosine, mental retardation associated with cerebral creatine deficiency ET, S-methyl-ET, trimethyl-histidine, OA, 2-ketobutyrate, and (33). Increased creatine might compensate for brain and muscle urate (Fig. 2A). Trimethyl-histidine and S-methyl-ET are in- dysfunction in frailty. Moreover, metabolites in UDP-glucuronate MEDICAL SCIENCES volved in ET synthesis, mainly in mushrooms and other fungi biosynthesis are much up-regulated. UDP-glucuronate is involved (20). OA is a tripeptide analog of (21), the precursor in the synthesis of ascorbic acid, formation of polysaccharides, and of which is 2-ketobutyrate. Thus, the ergothioneine and OA detoxification (34, 35). UDP-glucuronate is increased in frailty, pathways are greatly affected in frailty. Acetyl-carnosine, formed cognitive impairment, and low mobility, and UDP-glucose is up- from β-alanine and histidine, is enriched in muscle (22). Urate is regulated in cognitive impairment (Fig. 3D). It is noteworthy that one of the most abundant antioxidants in blood (23). Four of the antioxidant ET displays significant negative correlations with these seven antioxidants are also associated with cognitive im- creatine and UDP glucuronate (Pearson’s r = −0.76 and −0.56, pairment or low mobility: acetyl-carnosine, ET, OA, and respectively) (Fig. 3A and SI Appendix,Fig.S4). It is possible that 2-ketobutyrate (Fig. 2A). increased creatine or UDP-glucuronate compensates for decreased We observed significant decreases in five amino acids—me- antioxidative defense in frailty, which would make frailty reversible. thionine, proline, tryptophan, isoleucine, and leucine—in the Oxidative damage has been proposed to have a substantial impact frail subjects, while tryptophan, methionine, and proline were both on organismal aging in experimental models (36) and in ill- also reduced in patients manifesting cognitive impairment or low nesses of aging, such as Alzheimer’s disease (37). As such, declining mobility (Fig. 2B). Tryptophan is a precursor for the neuro- antioxidative defense could be involved in the pathogenesis of transmitters serotonin and dopamine and is involved in kynur- frailty, and oxidative stress may be a key vulnerability for frail enine metabolism in muscle (24), while leucine and isoleucine elderly people. While a total of 22 metabolites were identified are essential for maintaining muscle strength. According to relative to frailty, cognitive impairment, and hypomobility, Pearson’s correlation analysis, among the 15 metabolites asso- levels of 16 compounds—acetyl-carnosine, ET, S-methyl-ET, OA, ciated with frailty (Fig. 3A), five amino acids showed close cor- urate, methionine, proline, tryptophan, isoluecine, leucine, 1,5-AG, relations with several antioxidative metabolites. Correlation creatine, N3-methyl-histidine, isovaleryl-carnitine, hippurate, and coefficients of acetyl-carnosine with methionine and proline UDP-glucose—are abundant or moderately present in blood were r = 0.50 and 0.61, respectively, while those of S-methyl-ET (Fig. 4A), suggesting a possible involvement in pathogenesis. with methionine, proline, tryptophan, isoleucine, and leucine Among the 22 metabolites that we identified as relevant to were r = 0.52 to 0.65 (Fig. 3 A and B). Interestingly, methionine, frailty, cognitive /impairment, and hypomobility (Fig. 4A), partly proline, and tryptophan, the three amino acids that were de- overlapping but distinct metabolite profiles support the notion creased in frailty, have been reported as radical scavengers that frailty is an integrated spectrum of age-related disorders. in vitro (25, 26), consistent with recent findings in proteomic We addressed the question whether these metabolites are useful analysis (27). Thus, antioxidative defense is greatly impaired for diagnosis of frailty. Heatmap comparisons indicated distinct in frailty. distributions of 15 frailty markers between frail and nonfrail Twelve metabolites were identified as hypomobility markers groups (Fig. 4B). Similar results were observed among 6 me- (Fig. 3C and SI Appendix, Fig. S2B and Table S4), some of which tabolites for cognitive impairment and among 12 metabolites for are also frailty markers (acetyl-carnosine, OA, 2-ketobutyrate, low mobility (Fig. 4B). methionine, proline, and UDP-glucuronate) and cognitive markers Next, we applied principal component analysis (PCA) based (acetyl-carnosine and UDP-glucuronate) (Figs. 2 and 3D). Acetyl- on 10 metabolites related to EFS, MoCA-J, and TUG: acetyl- carnosine and UDP-glucuronate are linked to increased frailty and carnosine, ET, OA, methionine, proline, tryptophan, N3-methyl- to decreased cognition and mobility. The other six hypomobility histidine, creatine, UDP-glucuronate, and UDP-glucose. PCA markers include five decreased metabolites—N3-methyl-histidine, clearly distinguished frail elderly people from healthy counter- isovaleryl-carnitine, arginine, hippurate, and adenine (Fig. 3C and parts (Fig. 4C). Interestingly, nonfrail people with cognitive im- SI Appendix,Fig.S3B)—and increased N-acetyl-aspartate (Fig. 3D). pairment or hypomobility were also separated from the frail and

Kameda et al. PNAS | April 28, 2020 | vol. 117 | no. 17 | 9485 Downloaded by guest on September 24, 2021 Fig. 2. Antioxidants and amino acids were significantly decreased in frailty. (A) Seven antioxidants were identified as frailty markers: acetyl-carnosine, ET, S-methyl-ET, trimethyl-histidine, ophthalmic acid, 2-ketobutyrate, and urate. Decreased antioxidants were also observed in cognitive impairmentand hypomobility. (B) Five amino acids were decreased in frailty: isoleucine, leucine, methionine, tryptophan, and proline. Tryptophan was also decreased in cognitive impairment, while methionine and proline were also decreased in hypomobility. *P < 0.05. Error bars represent mean ± SD.

healthy groups. Analysis of a correlation network between EFS measured (19). In the EFS, a TUG test cutoff score of >10 s is classified as and 131 metabolites using Cytoscape 3.7.2 (38, 39) identified 17 hypomobility. metabolites significantly correlated with EFS, including 10 frailty The MoCA-J evaluates short-term memory, visuospatial ability, various SI Appendix executive functions, attention, concentration, working memory, language, markers ( , Table S2). Our metabolomic dissection of and temporal and spatial orientation. An MoCA-J score below a threshold of frailty markers suggests an involvement of antioxidation in cog- 25 to 26 (out of 30) is considered to indicate mild cognitive impairment nition and an involvement of nitrogen metabolism in mobility. (17, 18). These findings enhance our understanding of the pathogenesis of frailty and offer hope for interventions to maintain normal Blood Sample Preparation for Metabolomic Analysis. Preparation of human physiological levels of these metabolites. blood samples for metabolomic analysis has been described previously (9, 10, 40). In the morning, blood for clinical tests and metabolomic analysis was Materials and Methods drawn at the laboratory of Kyoto University Hospital. Until the time of blood sampling, all participants were requested not to have breakfast to ensure Clinical Assessment. All clinical data were collected at Kyoto University overnight fasting at least for 12 h, although they were allowed to spend Hospital. Patients who were bedridden or who had dysfunction (se- their time normally and to drink beverages without calories. Since some rum creatinine >2.0 mg/dL) or liver damage (serum aspartate aminotrans- metabolites are labile, blood samples were quickly quenched at −40 °C in > ferase and alanine aminotransferase 50 U/L), were excluded from the methanol to ensure quick sample processing. Then 10 nmol Hepes and Pipes study. Clinical interviews, physical examinations, and blood tests were per- were added to each sample to serve as internal standards. formed for 19 elderly participants. The EFS is an efficient diagnostic tool comprising 10 domains to assess cognitive ability (the clock-drawing test), LC-MS Conditions. Untargeted, comprehensive analysis by LC-MS was carried mobility (TUG test), and fundamental daily activity by questionnaire, in out as described previously (9, 10, 40). LC-MS data were obtained using an which a score ≥7 indicates frailty (range, 0 to 17) (16). The EFS elicits in- Ultimate 3000 DGP-3600RS liquid chromatograph and an LTQ Orbitrap mass formation on functional independence, including meal preparation, shop- spectrometer (Thermo Fisher Scientific). LC separation was done using a ZIC- ping, transportation, telephone use, housekeeping, laundry, money pHILIC column (Merck SeQuant; 150 mm × 2.1 mm, 5 μm particle size). The management, and medications. In the TUG test, the time it takes to stand up mobile phase was composed of ammonium carbonate buffer (10 mM, pH from a chair, walk normally to a point 3 m away, and return to sitting is 9.3) and acetonitrile. Gradient elution from 80 to 20% acetonitrile over

9486 | www.pnas.org/cgi/doi/10.1073/pnas.1920795117 Kameda et al. Downloaded by guest on September 24, 2021 MEDICAL SCIENCES

Fig. 3. Compounds relevant to muscle or nitrogen metabolism were decreased in hypomobility, while creatine and UDP-glucuronate were increased in frailty. (A) Pearson’s correlation analysis for 15 frailty markers. Positive and negative correlations are shown in red and blue, respectively. (B) Statistical analysis of the correlation between antioxidants and amino acids, acetyl-carnosine and proline (Upper) and S-methyl-ET and methionine (Lower). (C) Compounds involved in muscle or nitrogen metabolism declined in the low mobility group, including N3-methyl-histidine, isovaleryl-carnitine, arginine, and hippurate. (D) Four metabolites increased in frailty, cognitive impairment, or hypomobility. *P < 0.05; **P < 0.01. Error bars represent mean ± SD.

30 min at a flow rate of 100 μL/min was used. An electrospray ionization (ESI) software (http://www.r-project.org). Statistical analysis included Student’s t test source was used for MS detection. An injection of 1 μL was performed twice to confirm significant differences between groups (with statistical significance for each sample, once with the ESI in positive ionization mode and once with set at P < 0.05) and 95% confidence intervals, the ordinary least squares method the ESI in negative mode. Spray was set to 4.0 kV for positive ESI and 2.8 kV to confirm linear regression, and Pearson’s correlation to confirm relationships for negative ESI, while the capillary was adjusted to 350 or 300 °C. Nitrogen between metabolites and clinical data (assuming P < 0.05). PCA was used to gas was used as a carrier. The mass spectrometer was operated in full visualize the metabolomic model. A correlation network involving EFS and 131 scanning mode with a 100 to 1,000 m/z range and with MS/MS fragmenta- metabolites was analyzed using Cytoscape 3.7.2 (38) with Metscape (39). tion scanning in an automatic data-dependent manner. Data Availability. Raw LC-MS data in mzML format are available from the LC-MS Data Processing and Analysis. MZmine 2 (version 2.29) software (http:// MetaboLights repository, http://www.ebi.ac.uk/metabolights/ (accession no. mzmine.github.io) was used to measure peak areas for metabolites (41). MTBLS1540). Isotopic peaks were eliminated. Lists of peaks for individual samples were aligned according to their retention times and corresponding m/z values. A Ethics Statement. All participants signed informed consent forms prior to total of 131 nonselective metabolites were identified for each sample by examination, in accordance with the Declaration of Helsinki. Experiments comparing retention times and m/z values of peaks with those of standards were carried out in agreement with relevant rules and official guidelines in (Dataset S1) (9, 10, 40). If no standard compound was available, metabolites Japan. Approval for study protocols was given both by the Human Research were identified by the analysis of MS/MS spectra. Then all data acquired Ethics Committee of Kyoto University and by the Review Committee on were transferred into a spreadsheet, followed by analysis with R statistical Human Subjects Research at Okinawa Institute of Science and Technology.

Kameda et al. PNAS | April 28, 2020 | vol. 117 | no. 17 | 9487 Downloaded by guest on September 24, 2021 Fig. 4. Heatmap analysis and PCA for frailty. (A) Summary of metabolites related to frailty, cognitive impairment, and hypomobility. (B) Heatmap analysis of metabolites involved in frailty (Top), cognitive impairment (Middle), and hypomobility (Bottom). The heat map presents z-scores of peak areas from LC-MS analysis. (C) PCA plot of elderly subjects. Ten metabolites related to EFS, MoCA-J, and TUG were analyzed: acetyl-carnosine, ET, OA, methionine, proline, tryptophan, N3-methyl-histidine, creatine, UDP-glucuronate, and UDP-glucose.

ACKNOWLEDGMENTS. We thank Eri Shibata and Junko Takada for excellent (to M.Y.), and from the Ministry of Education, Culture, Sports, Science, and technical assistance and Dr. Steven D. Aird for editorial help. This work was Technology of Japan (to H.K.). The study was also generously supported by supported by grants from the Okinawa Institute of Science and Technology Okinawa Institute of Science and Technology Graduate University.

1. World Health Organization, World Report on Ageing and Health, (World Health 10. R. Chaleckis, I. Murakami, J. Takada, H. Kondoh, M. Yanagida, Individual variability in Organization, Geneva, Switzerland, 2015). human blood metabolites identifies age-related differences. Proc. Natl. Acad. Sci. 2. L. P. Fried et al.; Cardiovascular Health Study Collaborative Research Group, Frailty in U.S.A. 113, 4252–4259 (2016). older adults: Evidence for a phenotype. J. Gerontol. A Biol. Sci. Med. Sci. 56, 11. M. M. Marron et al., Metabolites associated with vigor to frailty among community- M146–M156 (2001). dwelling older black men. Metabolites 9, E83 (2019). 3. X. Chen, G. Mao, S. X. Leng, Frailty syndrome: An overview. Clin. Interv. Aging 9, 12. E. Pujos-Guillot et al., Identification of pre-frailty sub-phenotypes in elderly using 433–441 (2014). metabolomics. Front. Physiol. 9, 1903 (2019). 4. K. J. Ottenbacher et al., Mexican Americans and frailty: Findings from the Hispanic 13. N. J. W. Rattray et al., Metabolic dysregulation in vitamin E and carnitine shuttle Established Populations Epidemiologic Studies of the Elderly. Am. J. Public Health 99, energy mechanisms associate with human frailty. Nat. Commun. 10, 5027 (2019). 673–679 (2009). 14. G. Livshits et al., Multi-OMICS analyses of frailty and chronic widespread musculo- 5. K. Suhre et al.; CARDIoGRAM, Human metabolic individuality in biomedical and skeletal pain suggest involvement of shared neurological pathways. Pain 159, pharmaceutical research. Nature 477,54–60 (2011). 2565–2572 (2018). 6. J. van der Greef, H. van Wietmarschen, B. van Ommen, E. Verheij, Looking back into 15. K. Rockwood et al., A brief clinical instrument to classify frailty in elderly people. the future: 30 years of metabolomics at TNO. Mass Spectrom. Rev. 32, 399–415 (2013). Lancet 353, 205–206 (1999). 7. O. M. Wolkowitz et al., Leukocyte telomere length in major depression: Correlations 16. D. B. Rolfson, S. R. Majumdar, R. T. Tsuyuki, A. Tahir, K. Rockwood, Validity and re- with chronicity, inflammation and oxidative stress. Preliminary findings. PLoS One 6, liability of the Edmonton Frail Scale. Age Ageing 35, 526–529 (2006). e17837 (2011). 17. Y. Fujiwara et al., Brief screening tool for mild cognitive impairment in older Japa- 8. E. H. Blackburn, E. S. Epel, J. Lin, Human telomere biology: A contributory and nese: Validation of the Japanese version of the Montreal Cognitive Assessment. interactive factor in aging, disease risks, and protection. Science 350, 1193–1198 Geriatr. Gerontol. Int. 10, 225–232 (2010). (2015). 18. Z. S. Nasreddine et al., The Montreal Cognitive Assessment, MoCA: A brief screening 9. R. Chaleckis et al., Unexpected similarities between the Schizosaccharomyces and tool for mild cognitive impairment. J. Am. Geriatr. Soc. 53, 695–699 (2005). human blood metabolomes, and novel human metabolites. Mol. Biosyst. 10, 19. D. Podsiadlo, S. Richardson, The Timed “Up & Go”: A test of basic functional mobility 2538–2551 (2014). for frail elderly persons. J. Am. Geriatr. Soc. 39, 142–148 (1991).

9488 | www.pnas.org/cgi/doi/10.1073/pnas.1920795117 Kameda et al. Downloaded by guest on September 24, 2021 20. K. D. Asmus, R. V. Bensasson, J. L. Bernier, R. Houssin, E. J. Land, One-electron oxi- 30. M. S. Lustgarten, L. L. Price, A. Chale, E. M. Phillips, R. A. Fielding, Branched chain dation of ergothioneine and analogues investigated by pulse radiolysis: Redox re- amino acids are associated with muscle mass in functionally limited older adults. action involving ergothioneine and vitamin C. Biochem. J. 315, 625–629 (1996). J. Gerontol. A Biol. Sci. Med. Sci. 69, 717–724 (2014). 21. T. Soga et al., Differential metabolomics reveals ophthalmic acid as an oxidative stress 31. J. B. Walker, Creatine: Biosynthesis, regulation, and function. Adv. Enzymol. Relat. biomarker indicating hepatic glutathione consumption. J. Biol. Chem. 281, Areas Mol. Biol. 50, 177–242 (1979). 16768–16776 (2006). 32. S. Percário et al., Effects of creatine supplementation on oxidative stress profile of 22. A. Boldyrev, R. Song, D. Lawrence, D. O. Carpenter, Carnosine protects against ex- athletes. J. Int. Soc. Sports Nutr. 9, 56 (2012). citotoxic cell death independently of effects on . Neuroscience 33. S. Stöckler, F. Hanefeld, J. Frahm, Creatine replacement therapy in guanidinoacetate 94, 571–577 (1999). methyltransferase deficiency, a novel inborn error of metabolism. Lancet 348, – 23. R. El Ridi, H. Tallima, Physiological functions and pathogenic potential of uric acid: A 789 790 (1996). 34. R. H. Tukey, C. P. Strassburg, Human UDP-glucuronosyltransferases: Metabolism, ex- review. J. Adv. Res. 8, 487–493 (2017). pression, and disease. Annu. Rev. Pharmacol. Toxicol. 40, 581–616 (2000). 24. C. S. Katsanos, H. Kobayashi, M. Sheffield-Moore, A. Aarsland, R. R. Wolfe, A high 35. C. L. Linster, E. Van Schaftingen, Glucuronate, the precursor of vitamin C, is directly proportion of leucine is required for optimal stimulation of the rate of muscle protein formed from UDP-glucuronate in liver. FEBS J. 273, 1516–1527 (2006). synthesis by essential amino acids in the elderly. Am. J. Physiol. Endocrinol. Metab. 36. P. L. Larsen, Aging and resistance to oxidative damage in Caenorhabditis elegans. 291, E381–E387 (2006). Proc. Natl. Acad. Sci. U.S.A. 90, 8905–8909 (1993). 25. R. Marcuse, Antioxidative effect of amino-acids. Nature 186, 886–887 (1960). 37. W. R. Markesbery, J. M. Carney, Oxidative alterations in Alzheimer’s disease. Brain 26. X. Liang, L. Zhang, S. K. Natarajan, D. F. Becker, Proline mechanisms of stress survival. Pathol. 9, 133–146 (1999). – Antioxid. Redox Signal. 19, 998 1011 (2013). 38. R. Saito et al., A travel guide to Cytoscape plugins. Nat. Methods 9, 1069–1076 (2012). 27. S. D. Maleknia, M. Brenowitz, M. R. Chance, Millisecond radiolytic modification of 39. S. Basu et al., Sparse network modeling and metscape-based visualization methods peptides by synchrotron X-rays identified by mass spectrometry. Anal. Chem. 71, for the analysis of large-scale metabolomics data. Bioinformatics 33, 1545–1553 – 3965 3973 (1999). (2017). 28. J. Sjölin, H. Stjernström, S. Henneberg, L. Hambraeus, G. Friman, Evaluation of urinary 40. T. Teruya, R. Chaleckis, J. Takada, M. Yanagida, H. Kondoh, Diverse metabolic reac- 3-methylhistidine excretion in infection by measurements of 1-methylhistidine and tions activated during 58-hr fasting are revealed by non-targeted metabolomic the creatinine ratios. Am. J. Clin. Nutr. 49,62–70 (1989). analysis of human blood. Sci. Rep. 9, 854 (2019). 29. H. J. Lees, J. R. Swann, I. D. Wilson, J. K. Nicholson, E. Holmes, Hippurate: The natural 41. T. Pluskal, S. Castillo, A. Villar-Briones, M. Oresic, MZmine 2: Modular framework for history of a mammalian-microbial cometabolite. J. Proteome Res. 12, 1527–1546 processing, visualizing, and analyzing mass spectrometry-based molecular profile (2013). data. BMC Bioinformatics 11, 395 (2010). MEDICAL SCIENCES

Kameda et al. PNAS | April 28, 2020 | vol. 117 | no. 17 | 9489 Downloaded by guest on September 24, 2021